promptstringrequiredFull description of the picture.
Maximum 2500 characters; a prompt that does not describe a picture is rejected. Do not put pixel dimensions in the prompt, use width and height for that.
Logospell · Docs
Illustration tool
From one prompt, generate_illustration composes a single picture, a scene, banner, hero image, or portrait. The output lands at one of a fixed set of pixel sizes.
01 / 04 · Parameters
Single scenes
generate_illustration
One composed picture, a scene, banner, or hero image, rendered at one of the supported pixel sizes. 1 credit per call.
widthintegerrequiredOutput width in pixels. Must pair with height to form one of the supported sizes listed below.
heightintegerrequiredOutput height in pixels. Must pair with width to form one of the supported sizes listed below.
Supported output sizes, grouped by aspect ratio. Pick a shape, then pass the exact
width and height of one of its sizes; the ratio name is only a label, and any pair not listed is rejected. Largest first in each row, the rest are exact proportional downscales.Landscape, square and portrait shapes, each at its largest size and exact downscales of it. The full table of accepted width and height pairs appears here, and the
generate_illustration tool description carries the same list.formatstringoptionalpng is always lossless. jpg and webp use the quality parameter.
qualityintegeroptionalCompression quality for jpg and webp, from 1 (smallest file) to 100 (best quality). Ignored for png.
filenamestringoptionalName for the image file inside the returned zip.
unpackTostringoptionalLocal filesystem path your MCP client should extract the returned zip into.
Pass an absolute path. If omitted, the client picks a destination.
02 / 04 · Delivery
What lands on your filesystem
One picture
ready to use
Your AI client receives a short-lived download URL, fetches the ZIP, and extracts it into the path you passed as unpackTo. Set filename to control what the picture is called on disk; omit it and the file is named illustration.<format>.
After the client unzips the result you get exactly this on disk:
your filesystem
~/my-project/art/
├── scribe.webp
├── prompt.json (the request, for your records)
└── index.html (preview viewer)The index.html is a small preview page showing the picture with its prompt; open it in a browser for a quick visual check. If a connection ever drops mid-generation, list_recent_generations recovers the result.
03 / 04 · Prompt examples
How to write a good prompt
The prompt is the whole picture.
There is no separate style field here, so put subject, setting, lighting, mood, art treatment, and composition all in the one prompt. Each example below is complete and paste-ready.
prompt
Cinematic dusk
prompt
A lone lighthouse on a rocky headland at the last minute of dusk, beam cutting through low sea mist. Cold blue twilight in the sky against the warm sodium glow of the lamp, long wet reflections on the rocks. Wide cinematic shot, anamorphic feel, shallow horizon, photoreal with a faint film grain.Pick the exact minute of light, last minute of dusk reads warmer and moodier than just evening.
prompt
Editorial hero banner
prompt
Editorial hero banner for a fintech article: a single brushed-aluminum credit card standing on edge, casting a crisp shadow across a wide expanse of soft pearl-grey paper. Bright even studio light from the upper left, generous negative space on the right for a headline. Minimalist, high-key, premium product photography.Ask for negative space on a named side when you plan to lay text over the picture later.
prompt
Storybook interior
prompt
A cozy wizard's study at night, viewed from the doorway: floor-to-ceiling shelves of glowing potion bottles, a sleeping cat curled on a fat spellbook, steam rising from a copper kettle. Warm candlelight pooling in the center, deep blue shadows in the corners. Hand-painted storybook illustration, soft edges, rich saturated color.Anchor the camera with a phrase like viewed from the doorway so the model commits to one vantage point.
prompt
Character portrait
prompt
Three-quarter portrait of a weathered desert cartographer in her sixties, sun-creased face, brass goggles pushed up on her forehead, a rolled map under one arm. Soft window light from camera left, dark neutral background falling into shadow. Painterly realism in the spirit of Rembrandt lighting, shallow depth of field, focus on the eyes.For a person, name the age, the lighting direction, and where the focus lands, vague portraits drift generic.
prompt
Retro travel poster
prompt
Vintage 1950s travel poster of a cable car climbing a steep coastal city street toward a turquoise bay. Flat blocks of coral, teal, and cream with crisp hard-edged shapes, limited four-color palette, gentle paper texture. Bold diagonal composition leading the eye uphill, screen-print look. No text.Cap the palette out loud, limited four-color palette keeps a poster from muddying into realism.
prompt
Sci-fi matte painting
prompt
A vast colony greenhouse under a cracked Martian dome at dawn, rows of crops glowing under grow-lights while the rust-orange planet stretches beyond the glass. Two small figures walking a central path give the space its scale. Cinematic concept-art matte painting, volumetric god rays, cool interior light against warm exterior haze, ultra-wide establishing shot, fine atmospheric detail.Drop a small human figure into a vast scene and the model uses it to anchor the scale of everything else.
04 / 04 · Tips
Things worth knowing
Small habits that get you a stronger image.
One picture, not a set
generate_illustration composes a single complete scene: a hero image, a banner, an environment, or a portrait. If you want many separate isolated subjects sharing a style, reach for generate_image_set or generate_transparent_image_set instead.
Write the whole picture in the prompt
Subject, setting, lighting, mood, palette, art treatment, and composition all live in the one prompt, so put the whole picture in it.
Put dimensions in width and height, never the prompt
Set the shape with the width and height arguments and leave pixel counts out of the text. Repeating those counts in the prompt wastes characters and confuses the composition; the size arguments already handle it.
Only the listed size pairs work
width and height must be an exact pair from the supported sizes list; any other pair is rejected. The first pair in each row is that shape's largest size, and the rest are exact downscales of it.
Match the shape to the use
A banner wants a wide shape, a phone wallpaper a tall one, and a square avatar a square. Pick the aspect that fits, then a resolution from that row in the supported sizes list.
Name the light and the time of day
Lighting carries the mood of a single image. Say cold blue twilight or a single shaft of warm light from the right rather than just nice lighting, and the composition snaps into place.
Direct the camera and the framing
Tell the model where to stand. Cues like wide cinematic shot, three-quarter portrait, or viewed from the doorway fix the vantage point and stop the scene from drifting between compositions.
Pick the format, then name the output
format defaults to png; choose jpg or webp for smaller files where lossy is fine, and set quality (1-100, default 90) for those two (it is ignored for png). Each call costs 1 credit and returns a zip with the image, a prompt.json, and an index.html viewer; use filename and unpackTo to control the name and where it lands.